Zbornik 20. mednarodne multikonference

INFORMACIJSKA DRUŽBA – IS 2017 Zvezek A

Proceedings of the 20th International Multiconference

INFORMATION SOCIETY – IS 2017 Volume A

Slovenska konferenca o umetni inteligenci Slovenian Conference on Artificial Intelligence

Uredili / Edited by

Mitja Luštrek, Rok Piltaver, Matjaž Gams

http://is.ijs.si

9. - 13. oktober 2017 / 9th – 13th October 2017Ljubljana, Slovenia

Designing a Personal Decision Support System for

Congestive Heart Failure Management Marko Bohanec, Erik Dovgan, Pavel Maslov, Aljoša Vodopija,

Mitja Luštrek Jožef Stefan Institute

Department of Intelligent Systems, Department of Knowledge

Technologies Jamova cesta 39, 1000 Ljubljana,

Slovenia

{marko.bohanec, erik.dovgan, pavel.maslov, aljosa.vodopija,

mitja.lustrek}@ijs.si

Paolo Emilio Puddu, Michele Schiariti,

Maria Costanza Ciancarelli Sapienza University of Rome

Department of Cardiovascular, Respiratory, Nephrologic and Geriatric

Sciences Piazzale Aldo Moro 5, Roma 00185,

Italy

{paoloemilio.puddu. michele.schiariti}@uniroma1.it

mcostanza.ciancarelli@gmail.com

Anneleen Baert, Sofie Pardaens,

Els Clays Ghent University

Department of Public Health De Pintelaan 185 – 4K3, 9000 Gent,

Belgium

{anneleen.baert, sofie.pardaens,

els.clays}@ugent.be

ABSTRACT

In this paper, we describe the design of the HeartMan Decision

Support System (DSS). The DSS is aimed at helping patients

suffering from congestive heart failure to better manage their

disease. The support includes regular measurements of patients’

physical and psychological state using a wristband and mobile

device, and providing advice about physical exercise, nutrition,

medication therapy, and environment management. In the paper,

an overall architecture of the DSS is presented, followed by a

more detailed description of the module for physical exercise

management.

Categories and Subject Descriptors

H.4.2 [Types of Systems]: Decision support.

J.3 [Life and Medical Sciences]: Medical information systems.

General Terms

Algorithms, Management, Measurement, Design, Human Factors

Keywords

Decision Support System, Personal Health System, Congestive

Heart Failure, Physical Exercise, Decision Models

1. INTRODUCTION Congestive heart failure (CHF) occurs when the heart is unable to

pump sufficiently to maintain blood flow to meet the body's needs

[1]. Symptoms include shortness of breath, excessive tiredness,

and leg swelling. CHF is a common, chronic, costly, and

potentially fatal condition [2]. In 2015 it affected about 40 million

people globally. In developed countries, around 2% of adults have

heart failure, increasing to 6–10% in age over 65.

HeartMan (http://heartman-project.eu/) is a research project

funded by the European Union’s Horizon 2020 research and

innovation programme. The project aims to develop a personal

health system to help CHF patients manage their disease. CHF

patients have to take various medications, monitor their weight,

exercise appropriately, watch what they eat and drink, and make

other changes to their lifestyle. The HeartMan system will provide

accurate advice on disease management adapted to each patient in

a friendly and supportive fashion. The DSS follows the best

medical practices [3] and is designed so that it never suggests

anything that would harm the patient.

In this paper, we present the design of the HeartMan Decision

Support System (DSS), which was finalised in June 2017 [3]. In

section 2, we describe the overall functionality and architecture of

the system, and define the roles of its modules that address (1)

physiological measurements, (2) physical exercise, (3) nutrition,

(4) medication, (5) environment management, and (6)

management of calendars and plans. In section 3, we focus on the

physical exercise module and present its most important

components for (1) patients’ physical capacity assessment, (2)

weekly exercise planning, and (3) daily exercise management.

2. DESIGN OF THE HEARTMAN DSS The HeartMan DSS aims at providing medical advice to CHF

patients using predictive models, clinical care guidelines and

expert knowledge. The purpose of a typical DSS is to passively

present information to decision makers so that they can make

maximally informed decisions. This DSS, however, is intended

for patients who have limited medical knowledge and are

consequently expected to follow guidelines with little discretion.

Because of that, the DSS actively provides advice to patients,

although it does offer choice where appropriate. In this way, it

belongs to the category of cooperative DSS [4].

Phone sensors

Health devices

Monitoring physical and psychological

state

Mobile application

Decision Support Modules

1. Physiological measurements

2. Physical exercise3. Nutrition4. Medication5. Environment

management6. Calendars and plans

Data management

Web interface

Figure 1: Overall architecture of the HeartMan DSS.

67

The overall architecture of the HeartMan DSS is shown in Figure

1. The system will use wrist-band sensors to monitor patient’s

physical activity, heart rate and some other physiological signs. In

addition, it will receive data from additional devices, such as

scales, smartphone and from the patient via the user interface of

the mobile application. This will allow the system to identify the

patient's current physical and psychological characteristics. This

data will be combined with patient’s health data to help them

decide on disease control measures. The advice will be tailored to

the patient's medical condition by adapting it to the patient's

psychological profile (such as normal, poorly motivated,

depressed, and anxious) and current health state. The advice will

be shown at the mobile app. A web-based interface will be

provided to the physician, too, who will be able to monitor the

patient’s health state and progress, and define or approve

parameters that affect the advice given to the patient.

The core of the HeartMan DSS are modules that interpret

patient’s data and make recommendations. There are six main

modules, which address the following aspects of health

management:

1. Physiological measurements: CHF patients should perform

various physiological measurements on a regular basis, such

as measuring their weight, blood pressure, heart rate, etc. This

raises the patients’ awareness of their health, provides

valuable information to their physicians, and provides inputs

to the DSS. For this purpose, the DSS reminds the patient to

regularly perform these measurements, provides the

functionality to carry them out, and manage the collected data.

2. Exercise: Physical conditioning by exercise training reduces

mortality and hospitalization, and improves exercise tolerance

and health-related quality of life. For this purpose, the DSS

provides a comprehensive exercise programme, which is

detailed later in section 3.

3. Nutrition: CHF patients should maintain their body weight

and take care of their diet, for instance, not eating too much

salt or drinking too much fluid. The DSS assesses the

patients’ nutrition behavior, educates them through a quiz and

provides advice towards a healthy diet.

4. Medication: Good adherence to medication therapy decreases

mortality and morbidity, and improves well-being of CHF

patients. For this purpose, the DSS reminds the patient to take

medications and assesses the patient’s adherence to the

medication scheme. For each medication, the patient may

obtain an explanation why the adherence is important.

5. Environment management: Environmental conditions, such as

temperature and humidity, may affect the patient’s feeling of

health. Combining both, the patient’s and environmental

conditions, the DSS advises the patient how to change the

environment to improve their health feeling.

6. Calendars and plans: Given all the DSS aspects

(measurement, exercise, nutrition, medication and

environment management) and many interactions between

them, it is important to sensibly arrange all the activities and

notifications, for instance not sending nutrition advice during

exercise or suggesting physical exercise after taking diuretics.

This DSS module thus coordinates the activities and arranges

all the plans into one single calendar.

To date, all these modules have been designed in terms of their

functionality, requirements, input data, processing, outputs, and

distribution between the client (patient’s mobile app) and the

server (the DSS in “the cloud”) [3].

3. PHYSICAL EXERCISE MODULE The HeartMan DSS administers a comprehensive exercise

programme. At the beginning, the DSS collects medical

information and assesses patient’s physical capacity in order to

plan the difficulty level of the exercises. Then, the DSS provides a

weekly set of endurance and resistance exercises, which increase

in difficulty as the patient becomes fitter. The DSS also guides the

patient during each exercise session: it checks whether the patient

is ready to start, then provides instructions, and finally asks the

patient to evaluate the exercise. The exercise module follows the

guidelines provided in [5] with minor modifications to fit in a

mobile application.

3.1 Physical Capacity Assessment Prior to starting using the HeartMan DSS, the patients should

perform a cardiopulmonary exercise (cycloergometry) test to

assess their physical capacity. Alternatively, when using the

system in a supervised, standardized setting, patients can perform

a 6-minute walking test. On this basis, the physical capacity of

each patient is assessed as “low” (less than 1 W/kg measured by

cycloergometry or less than 300 m walked in 6 minutes) or

“normal” (otherwise).

3.2 Weekly Exercise Planning The DSS system provides the patient with a combined endurance

and resistance exercise programme. Both types follow the same

principle described with four parameters: frequency (times per

week), intensity, duration and type. These parameters are

combined with the physical capacity to make a weekly exercise

plan for each patient. For instance, low-capacity patients start with

very light 10-15-minute endurance exercises twice per week.

According to the patient’s progress, these parameters may change

with time. In the HeartMan DSS, the progress is prescribed by

two models:

EnduranceFrequency: a model for suggesting weekly

frequency of endurance exercises;

EnduranceTime: a model for suggesting weekly time

boundaries of endurance exercises.

Both models are formulated using a qualitative multi-criteria

decision analysis method DEX [6]. Here, we illustrate the

approach describing the EnduranceFrequency model, whose

structure is shown in Figure 2.

DEXi EnduranceFrequency.dxi 15.5.17 Page 1

Scales Attribute Scale EnduranceFrequency 2x; 3x; 4x; 5x

Normative 2x; 3x; 4x; 5xCategory low; normalWeek 1; 2; 3; 4; 5; 6; 7; 8; 9; 10; 11; 12; 13; 14; 15; 16; 17; 18; 19; 20; 21; 22; 23; 24; more

Current 2x; 3x; 4x; 5xTransition decrease; stay; increase; automatic

MedicalAssessment decrease; stay; increase; automaticPatientsAssessment decrease; stay; increase; automatic

Figure 2: Structure of the EnduranceFrequency model.

The EnduranceFrequency model is aimed at suggesting the

frequency of exercises for the next week, based on the patient’s

physical capacity, week in the programme, current frequency, and

the possible physician’s and patient’s suggestions for the change.

In other words, the model takes into account both the normative

68

(as proposed by a general programme) and actual (as practiced by

the patient) frequency, leveraging the patient’s and physician’s

opinion about the suggestion for the subsequent week.

The overall recommendation, which is 2, 3, 4, or 5 times per

week, is represented by the root attribute EnduranceFrequency

(Figure 2). The recommendation depends on three sub-criteria:

1. Normative: Frequency as suggested according to the default

programme. It depends on the patient’s physical capacity

(“low” or “normal” Category) and the current Week. The

progression is defined by rules presented in Table 1.

2. Current: The frequency of exercises currently carried out by

the patient; it can run ahead or behind the Normative plan. In

order to make only small and gradual changes to the

frequency, Current is compared to Normative and only a one-

step change is suggested in each week.

3. Transition is an attribute that captures the patient’s wish and

the physician’s opinion about changing the frequency. The

possible values are “decrease”, “same”, “increase” or

“automatic”; the latter is meant to suggest the frequency

according to the normal plan, for instance, when neither the

patient or physician have given any suggestion. The patient’s

and physician’s suggestions are combined according to

decision rules shown in Table 2. The first two rules say that

whenever the patient or the physician suggest to decrease the

frequency, it should indeed be decreased (the symbol ‘*’

represents any possible value). Rules 3 and 4 suggest to keep

the current frequency whenever one of the participants

suggests so, unless the other participant suggests “decrease”

Rules 5 and 6 define a similar reasoning for “increase”. If

both participants have no particular suggestions, the

“automatic” transition according to the normal plan takes

place.

Table 1: Decision table defining the Normative frequency.

Category Week Normative

1 low <=4 2x

2 low 5–12 3x

3 normal <=6 3x

4 low 13–18 4x

5 normal 7–12 4x

6 low >=19 5x

7 normal >=13 5x

Table 2: Decision rules for Transition.

MedicalAssessment PatientsAssessment Transition

1 decrease * decrease

2 * decrease decrease

3 stay not decrease stay

4 not decrease stay stay

5 increase increase or automatic increase

6 increase or automatic increase increase

7 automatic automatic automatic

3.3 Daily Exercise Management Once a weekly plan has been established, the HeartMan DSS

assists the patient in carrying out their daily exercises. This

consists of four activities: (1) reminding the patient, (2) pre-

exercise checking, (3) exercise monitoring, and (4) post-exercise

assessment.

3.3.1 Reminding the patient Patients can choose the days when they want to exercise (e.g.,

every Tuesday, Thursday and Sunday). On these particular days,

the patients are at some predefined morning time reminded about

the daily exercise. Another reminder is issued if the exercise has

not been completed before the given afternoon time.

3.3.2 Before the exercise Before the start of each exercise session, the HeartMan DSS

checks if all prior-exercise requirements are met, and advises the

patients about safety. Figure 3 shows the decision model.

Figure 3: Pre-exercise assessment.

1. Information requirements: The blood pressure should have

been measured during the day. If not, the patients are

instructed to measure it. The pre-exercise heart rate is

measured automatically by the wristband; the system makes

sure that it is actually worn.

2. Medication requirements: Patients are asked to fill a check-list

of frequently seen side effects based on their medication

schemes and symptoms (e.g., dizziness and chest pain). On

this basis the DSS checks for any possible restrictions due to

medications or symptoms and suggests rescheduling the

session if necessary. The physician or nurse are contacted if

severe side effects are present. In the case of dizziness or chest

pain, patients are instructed to rest until the symptoms are no

longer present.

3. Physiological requirements: If all the requirements are met,

patients can start with the exercise, otherwise they are

instructed to repeat the measurements after five minutes of

rest. If after re-checking the measurements are still not within

safe limits, exercise is not allowed and patients are advised to

contact their physician or heart failure nurse.

Again, a DEX [3, 6] model is employed for assembling and

checking the medical conditions, which include medical intake,

comorbidities and current physical condition of the patient. The

structure and scales of the PreExerciseRequirements model are

shown in Figure 4. All attributes are binary (“yes”/“no” or

“not_met”/“met”). The values of the input attributes are

69

determined from patient data whenever the pre-exercise

requirements are checked (normally once per day before making

exercises). The subtrees of the model comprise four main groups

of possible reasons against participating in the exercises:

Blood coagulation: Whenever the patient takes anticoagulants

and there are symptoms indicating a possible bleeding: rash,

hemorrhages, or neurological symptoms.

Medication intake: Whenever one of the following

medications has been taken 2 hours or less before the

exercise: beta blockers, ACE inhibitors, ARBs, diuretics, or

loop diuretics.

Heart rate: Whenever the patient takes Digitalis and his/her

HR is less than 45 bpm.

Blood pressure: Whenever there are risks of hypertension

(taking ACE inhibitors or ARBs, and the patient has persistent

low blood pressure or persistent cough) or problems regarding

the systolic blood pressure (when the patient’s systolic blood

pressure is less than 105 and he/she recently took loop

diuretics). DEXi PreExerciseRequirements.dxi 16.5.17 Page 1

Scales

Attribute Scale PreExerciseRequirements not_met; met

BloodCoagulationReasons yes; noTakesAnticoagulats yes; noPossibleBleeding yes; no

Rash yes; noHemorrhages yes; noNeurologicalSymptoms yes; no

MedicationIntakeReasons yes; noIntake<2hours yes; noExercisePreventionMedications yes; no

TakesBetaBlockers yes; noTakesACEInhibitors yes; noTakesARBs yes; noTakesDiuretics yes; noTakesLoopDiuretics yes; no

HeartRateReasons yes; noTakesDigitalis yes; noHR<45 yes; no

BloodPressureReasons yes; noHypertensionReasons yes; noTakesACEInhibitors yes; noTakesARBs yes; noPersistentLowBloodPressure yes; noPersistentCough yes; no

SystolicPressureReasons yes; noTakesLoopDiuretics yes; noTookLoopDiuretics yes; noSYS<105 yes; no

Figure 4: Structure of the PreExerciseRequirements model.

3.3.3 During the exercise If the exercise is allowed, a list of exercises is shown to the

patient, who can then select the preferred exercise. After selecting

the exercise, a detailed description (text or graphical) regarding

the exercise is provided.

During the exercise, the heart rate and systolic blood pressure are

continuously measured by the wristband. The patients are advised

to stop the exercise in case of symptoms or measurements lying

outside of prescribed safety margins. If the heart rate is within the

safety limits, but too low or too high, the patent is advised to

increase or decrease the intensity, respectively. The system also

advises the patients about the exercise duration and is capable of

recognizing a premature ending.

3.3.4 After the exercise After completing the exercise, the patients can rate their feeling of

intensity (very light, light, moderate, intense, very intense). Then

the system assesses the exercise based on measurements recorded

during the exercise. It checks if the exercise was prematurely

finished and if the intensity was on average in the prescribed

limits. The system takes into account this information when

assessing the adherence to the exercise plan and the patient’s

improvement. Independent of this, the exercise is shown as

completed and the weekly plan is updated.

4. CONCLUSIONThis paper described the design of the HeartMan DSS that is

concerned with “medical” interventions (i.e., interventions that try

to improve the patients’ physical condition as opposed to

psychological). The DSS is based on clinical guidelines for the

self-management of CHF, additional medical literature and expert

knowledge from the project consortium. The DSS is designed in

terms of process models (the order of actions and questions) and

decision models (how to make some complex decisions –

branching in the process models) for five main topics of CHF

management: physiological measurements, exercise, nutrition,

medication, and environment management.

The DSS is currently being integrated in the overall HeartMan

platform. A comprehensive validation involving 120 CHF patients

(of whom 40 are controls without the DSS) is planned for 2018.

5. ACKNOWLEDGMENTSThe HeartMan project has received funding from the European

Union’s Horizon 2020 research and innovation programme under

grant agreement No 689660. Project partners are Jožef Stefan

Institute, Sapienza University, Ghent University, National

Research Council, ATOS Spain SA, SenLab, KU Leuven, MEGA

Electronics Ltd and European Heart Network.

6. REFERENCES[1] Heart failure. Health Information. Mayo Clinic. 23

December 2009. DS00061. http://www.mayoclinic.org/

diseases-conditions/heart-failure/basics/definition/con-

20029801.

[2] McMurray, J.J., Pfeffer. M.A. 2005. Heart failure. Lancet

365 (9474): 1877–89. DOI=

https://doi.org/10.1016%2FS0140-6736%2805%2966621-4.

[3] Bohanec, M., Vodopija, A., Dovgan, W., Maslov, P.,

Luštrek, M., Puddu, P.E., Clays, E., Pardaens, S., Baert, A.

(2017). Medical DSS for CHF. HeartMan Deliverable D4.1.

[4] Turban, E., Sharda, R., Delen, D., King, D. (2010). Business

Intelligence. 2nd Edition, Prentice Hall.

[5] Piepoli, M.F., Conraads, V., Corrà, U., et al. (2011):

Exercise training in heart failure: from theory to practice. A

consensus document of the Heart Failure Association and the

European Association for Cardiovascular Prevention and

Rehabilitation. European Journal of Heart Failure 13, 347–

357.

[6] Bohanec, M., Rajkovič, V., Bratko, I., Zupan, B., Žnidaršič,

M. (2013): DEX methodology: Three decades of qualitative

multi-attribute modelling. Informatica 37, 49–5

70